Remove underscore prefixes from compiler.h macros

Differential Revision: D61346624

Pull Request resolved: #4737

Author: dbort

backends/apple/coreml/runtime/test/CoreMLBackendDelegateTests.mm

@@ -32,7 +32,7 @@
     {}
 
     Result<FreeableBuffer> load(
-        size_t offset, size_t size, __ET_UNUSED const DataLoader::SegmentInfo& segment_info) const override {
+        size_t offset, size_t size, ET_UNUSED const DataLoader::SegmentInfo& segment_info) const override {
         NSData *subdata = [data_ subdataWithRange:NSMakeRange(offset, size)];
         return FreeableBuffer(subdata.bytes, size, nullptr);
     }

backends/apple/mps/runtime/MPSBackend.mm

@@ -55,7 +55,7 @@ bool is_available() const override {
 
   // Function that actually executes the model in the backend.
   Error execute(
-    __ET_UNUSED BackendExecutionContext& context,
+    ET_UNUSED BackendExecutionContext& context,
     DelegateHandle* handle,
     EValue** args) const override {
     auto executor = static_cast<mps::delegate::MPSExecutor*>(handle);

backends/apple/mps/runtime/MPSCompiler.h

@@ -24,7 +24,7 @@ class MPSCompiler {
   // Takes Flatbuffer Serialized MPS Model and rebuilds the MPSGraphExecutable
   // returns an executor object that holds the MPS runtime object which we
   // can then use to set inputs and run inference using the MPSGraphExecutable.
-  __ET_NODISCARD static Error compileModel(
+  ET_NODISCARD static Error compileModel(
       const void* buffer_pointer,
       size_t num_bytes,
       MPSExecutor* executor,

backends/apple/mps/runtime/MPSCompiler.mm

@@ -32,7 +32,7 @@
 Builds the mps runtime object using the buffer pointer. The buffer pointer
 must be a valid pointer to the serialized mps object.
 */
-__ET_NODISCARD Error MPSCompiler::compileModel(
+ET_NODISCARD Error MPSCompiler::compileModel(
   const void* buffer_pointer,
   size_t num_bytes,
   MPSExecutor* executor,

backends/apple/mps/runtime/MPSExecutor.h

@@ -73,9 +73,9 @@ class MPSExecutor {
     return _executable;
   }
 
-  __ET_NODISCARD Error forward(std::vector<const Tensor*>& outputs);
+  ET_NODISCARD Error forward(std::vector<const Tensor*>& outputs);
 
-  __ET_NODISCARD Error
+  ET_NODISCARD Error
   set_inputs_outputs(std::vector<const Tensor*>& inputs, std::vector<const Tensor*>& outputs);
 
   Error initDataBuffers();

backends/apple/mps/runtime/MPSExecutor.mm

@@ -37,7 +37,7 @@ @interface MPSGraphExecutable()
   _outputsArray = [[NSMutableArray<MPSGraphTensorData *> alloc] initWithCapacity:getNumOutputs()];
 }
 
-__ET_NODISCARD Error
+ET_NODISCARD Error
 MPSExecutor::set_inputs_outputs(std::vector<const Tensor*>& inputs, std::vector<const Tensor*>& outputs) {
   ET_CHECK_OR_RETURN_ERROR(inputs.size() == getNumInputs(), Internal, "Inputs mismatch");
   ET_CHECK_OR_RETURN_ERROR(outputs.size() == getNumOutputs(), Internal, "Outputs mismatch");
@@ -61,7 +61,7 @@ @interface MPSGraphExecutable()
   return Error::Ok;
 }
 
-__ET_NODISCARD Error MPSExecutor::forward(std::vector<const Tensor*>& outputs) {
+ET_NODISCARD Error MPSExecutor::forward(std::vector<const Tensor*>& outputs) {
   Error err = Error::Ok;
   MPSStream* mpsStream = getDefaultMPSStream();
   if (mpsStream->commitAndContinueEnabled() || mpsStream->hasLiveCommandBuffer()) {

backends/apple/mps/runtime/MPSStream.h

@@ -63,7 +63,7 @@ class MPSStream {
   MPSCommandBuffer* commandBuffer();
   id<MTLComputeCommandEncoder> commandEncoder();
   void endKernelCoalescing();
-  __ET_NODISCARD Error synchronize(SyncType syncType);
+  ET_NODISCARD Error synchronize(SyncType syncType);
   bool commitAndContinueEnabled();
   void copy(
       id<MTLBuffer> srcBuffer,

backends/apple/mps/runtime/MPSStream.mm

@@ -55,7 +55,7 @@ @interface MPSGraphExecutionDescriptor ()
   return _commandEncoder;
 }
 
-__ET_NODISCARD
+ET_NODISCARD
 Error MPSStream::synchronize(SyncType syncType) {
   endKernelCoalescing();
   switch(syncType) {
@@ -157,7 +157,7 @@ @interface MPSGraphExecutionDescriptor ()
       endKernelCoalescing();
       if (@available(iOS 13.0, *)) {
         id<MTLBlitCommandEncoder> blitEncoder = [commandBuffer() blitCommandEncoder];
-        
+
         [blitEncoder copyFromBuffer:srcBuffer
                        sourceOffset:(NSUInteger)srcOffset
                            toBuffer:dstBuffer

backends/cadence/executor_runner.cpp

@@ -83,10 +83,10 @@ void et_pal_emit_log_message(
     et_timestamp_t timestamp,
     et_pal_log_level_t level,
     const char* filename,
-    __ET_UNUSED const char* function,
+    ET_UNUSED const char* function,
     size_t line,
     const char* message,
-    __ET_UNUSED size_t length) {
+    ET_UNUSED size_t length) {
   PRINTF("\r%s\n", message);
 }
 

backends/mediatek/runtime/NeuronBackend.cpp

@@ -67,7 +67,7 @@ Result<DelegateHandle*> NeuronBackend::init(
 }
 
 Error NeuronBackend::execute(
-    __ET_UNUSED BackendExecutionContext& context,
+    ET_UNUSED BackendExecutionContext& context,
     DelegateHandle* handle,
     EValue** args) const {
   NeuronExecuTorchDelegate* delegate =

backends/mediatek/runtime/include/NeuronBackend.h

@@ -34,7 +34,7 @@ class NeuronBackend final : public PyTorchBackendInterface {
       ArrayRef<CompileSpec> compile_specs) const override;
 
   Error execute(
-      __ET_UNUSED BackendExecutionContext& context,
+      ET_UNUSED BackendExecutionContext& context,
       DelegateHandle* handle,
       EValue** args) const override;
 
@@ -111,7 +111,7 @@ class NeuronExecuTorchDelegate {
     return NEURON_NO_ERROR;
   }
 
-  Error execute(__ET_UNUSED BackendExecutionContext& context, EValue** args)
+  Error execute(ET_UNUSED BackendExecutionContext& context, EValue** args)
       const;
 
  private:

backends/qualcomm/runtime/QnnExecuTorchBackend.h

@@ -24,7 +24,7 @@ class QnnExecuTorchBackend final : public PyTorchBackendInterface {
       ArrayRef<CompileSpec> compile_specs) const override;
 
   Error execute(
-      __ET_UNUSED BackendExecutionContext& context,
+      ET_UNUSED BackendExecutionContext& context,
       DelegateHandle* handle,
       EValue** args) const override;
 

backends/vulkan/runtime/VulkanBackend.cpp

@@ -421,7 +421,7 @@ class VulkanBackend final : public PyTorchBackendInterface {
     return true;
   }
 
-  __ET_NODISCARD Error
+  ET_NODISCARD Error
   compileModel(const void* buffer_pointer, ComputeGraph* compute_graph) const {
     Result<VulkanDelegateHeader> header =
         VulkanDelegateHeader::parse(buffer_pointer);
@@ -485,7 +485,7 @@ class VulkanBackend final : public PyTorchBackendInterface {
   }
 
   Error execute(
-      __ET_UNUSED BackendExecutionContext& context,
+      ET_UNUSED BackendExecutionContext& context,
       DelegateHandle* handle,
       EValue** args) const override {
     EXECUTORCH_SCOPE_PROF("VulkanBackend::execute");

backends/xnnpack/runtime/XNNCompiler.cpp

@@ -1608,7 +1608,7 @@ Builds the xnnpack runtime object using the buffer pointer. The buffer pointer
 must be a valid pointer to the serialized xnnpack object. It also fills the
 XNNExecutor object with the built xnn_runtime and the input/output ids.
 */
-__ET_NODISCARD Error XNNCompiler::compileModel(
+ET_NODISCARD Error XNNCompiler::compileModel(
     const void* buffer_pointer,
     size_t num_bytes,
     XNNExecutor* executor,

backends/xnnpack/runtime/XNNCompiler.h

@@ -25,7 +25,7 @@ class XNNCompiler {
   // Takes Flatbuffer Serialized XNNPACK Model and rebuilds the xnn-subgraph
   // returns an executor object that holds the xnn runtime object which we
   // can then use to set inputs and run inference using the xnn graph.
-  __ET_NODISCARD static Error compileModel(
+  ET_NODISCARD static Error compileModel(
       const void* buffer_pointer,
       size_t num_bytes,
       XNNExecutor* executor,

backends/xnnpack/runtime/XNNExecutor.cpp

@@ -22,7 +22,7 @@ using SizesType = exec_aten::SizesType;
  * inputs/outputs externals_ is resized to the total number of inputs and
  * outputs
  */
-__ET_NODISCARD Error XNNExecutor::initialize(
+ET_NODISCARD Error XNNExecutor::initialize(
     xnn_runtime_t runtime,
     std::vector<uint32_t>&& input_ids,
     std::vector<uint32_t>&& output_ids) {
@@ -62,7 +62,7 @@ __ET_NODISCARD Error XNNExecutor::initialize(
  * runtime correspond to their index in the list of arg passed into
  * delegate->execute()
  */
-__ET_NODISCARD Error XNNExecutor::prepare_args(EValue** args) {
+ET_NODISCARD Error XNNExecutor::prepare_args(EValue** args) {
   // Create xnn_externals_value from evalue args
   xnn_status status;
   for (uint32_t i = 0; i < externals_.size(); ++i) {
@@ -128,7 +128,7 @@ __ET_NODISCARD Error XNNExecutor::prepare_args(EValue** args) {
  * We first setup the runtime by feeding the externals_ to runtime setup.
  * After which we then execute the runtime through invoke_runtime.
  */
-__ET_NODISCARD Error XNNExecutor::forward(BackendExecutionContext& context) {
+ET_NODISCARD Error XNNExecutor::forward(BackendExecutionContext& context) {
   ET_CHECK_OR_RETURN_ERROR(
       runtime_ != nullptr,
       Internal,
@@ -180,7 +180,7 @@ __ET_NODISCARD Error XNNExecutor::forward(BackendExecutionContext& context) {
  * XNNPACK gives the index tensor to us as int32, we need to convert it
  * back to int64 for ExecuTorch.
  */
-__ET_NODISCARD Error XNNExecutor::resize_outputs(EValue** args) const {
+ET_NODISCARD Error XNNExecutor::resize_outputs(EValue** args) const {
   size_t output_idx_start = input_ids_.size();
   for (size_t i = output_idx_start; i < externals_.size(); ++i) {
     uint32_t ext_id = externals_[i].id;

backends/xnnpack/runtime/XNNExecutor.h

@@ -51,7 +51,7 @@ class XNNExecutor {
    * The input/output ids are expected to be sorted in order of their
    * flatbuffer id_outs
    */
-  __ET_NODISCARD Error initialize(
+  ET_NODISCARD Error initialize(
       xnn_runtime_t runtime,
       std::vector<uint32_t>&& input_ids,
       std::vector<uint32_t>&& output_ids);
@@ -62,19 +62,19 @@ class XNNExecutor {
    * input shapes will be propagated through the runtime, and perform
    * any additional memory planning as needed
    */
-  __ET_NODISCARD Error prepare_args(EValue** args);
+  ET_NODISCARD Error prepare_args(EValue** args);
 
   /**
    * Executes the graph using the args prepared at prepare_args().
    */
-  __ET_NODISCARD Error forward(BackendExecutionContext& context);
+  ET_NODISCARD Error forward(BackendExecutionContext& context);
 
   /**
    * Prepares the outputs to be returned by the delegate
    *
    * Performs any post processing of outputs like tensor resizing
    */
-  __ET_NODISCARD Error resize_outputs(EValue** args) const;
+  ET_NODISCARD Error resize_outputs(EValue** args) const;
 
   friend class XNNCompiler;
 };

backends/xnnpack/threadpool/threadpool.cpp

@@ -117,7 +117,7 @@ ThreadPool* get_threadpool() {
   // @lint-ignore CLANGTIDY facebook-hte-std::call_once
   std::call_once(
       flag, []() { pthread_atfork(nullptr, nullptr, child_atfork); });
-  if __ET_UNLIKELY (leak_corrupted_threadpool) {
+  if ET_UNLIKELY (leak_corrupted_threadpool) {
     leak_corrupted_threadpool = false;
     if (auto leaked = threadpool.release()) {
       auto t = leaked->get_thread_count();

docs/Makefile

@@ -2,7 +2,7 @@
 #
 
 # You can set these variables from the command line.
-# SPHINXOPTS    = -WT --keep-going TODO(T165752164) fix sphinx warnings around preprocess macros in cpp like __ET_DEPRECATED
+# SPHINXOPTS    = -WT --keep-going TODO(T165752164) fix sphinx warnings around preprocess macros in cpp like ET_DEPRECATED
 SPHINXBUILD   = sphinx-build
 SPHINXPROJ    = ExecuTorch
 SOURCEDIR     = source

docs/source/api-life-cycle.md

@@ -113,11 +113,11 @@ usage</a>
    </td>
    <td>
 
-Use <code>__ET_DEPRECATED</code> macros. See <a href="https://github.com/pytorch/executorch/blob/8e0f856ee269b319ac4195509cf31e3f548aa0e8/runtime/executor/program.h#L81">example usage</a>
+Use <code>ET_DEPRECATED</code> macros. See <a href="https://github.com/pytorch/executorch/blob/8e0f856ee269b319ac4195509cf31e3f548aa0e8/runtime/executor/program.h#L81">example usage</a>
 
 <p>
 <p>
-Use <code>__ET_EXPERIMENTAL</code> macros (TODO not yet implemented)
+Use <code>ET_EXPERIMENTAL</code> macros (TODO not yet implemented)
 </ul>
    </td>
    <td>

docs/source/compiler-delegate-and-partitioner.md

@@ -87,22 +87,22 @@ function which will be called when the program is out of its lifespan.
 
 ```cpp
 // Runtime check
-__ET_NODISCARD bool is_available();
+ET_NODISCARD bool is_available();
 
 // Runtime initialization
-__ET_NODISCARD virtual Result<DelegateHandle*> init(
+ET_NODISCARD virtual Result<DelegateHandle*> init(
     BackendInitContext& context,
     FreeableBuffer* processed,
     ArrayRef<CompileSpec> compile_specs);
 
 // Runtime execution
-__ET_NODISCARD virtual Error execute(
+ET_NODISCARD virtual Error execute(
     BackendExecutionContext& context,
     DelegateHandle* handle,
     EValue** args);
 
 // [optional] Runtime destroy. Destroy the resource held by the backend
-virtual void destroy(__ET_UNUSED DelegateHandle* handle);
+virtual void destroy(ET_UNUSED DelegateHandle* handle);
 ```
 
 The diagram looks like following
@@ -114,7 +114,7 @@ The diagram looks like following
 
 In order to make backend available to ExecuTorch runtime, it must be registered via the `register_backend` API:
 ```cpp
-__ET_NODISCARD Error register_backend(const Backend& backend);
+ET_NODISCARD Error register_backend(const Backend& backend);
 ```
 
 Static registeration, i.e., at libraray init or load time, of a backend can be achieved as follows:

docs/website/docs/tutorials/bundled_program.md

@@ -88,7 +88,7 @@ To execute the program on the bundled input, we need to load the bundled input i
  * @returns Return Error::Ok if load successfully, or the error happens during
  * execution.
  */
-__ET_NODISCARD Error LoadBundledInput(
+ET_NODISCARD Error LoadBundledInput(
     Method& method,
     serialized_bundled_program* bundled_program_ptr,
     size_t testset_idx);
@@ -111,7 +111,7 @@ We call `torch::executor::bundled_program::VerifyResultWithBundledExpectedOutput
  * @returns Return Error::Ok if two outputs match, or the error happens during
  * execution.
  */
-__ET_NODISCARD Error VerifyResultWithBundledExpectedOutput(
+ET_NODISCARD Error VerifyResultWithBundledExpectedOutput(
     Method& method,
     serialized_bundled_program* bundled_program_ptr,
     size_t testset_idx,

examples/apple/coreml/executor_runner/main.mm

@@ -160,7 +160,7 @@ Args parse_command_line_args(NSArray<NSString *> *args) {
     :data_(read_data(filePath))
     {}
 
-    Result<FreeableBuffer> load(size_t offset, size_t size, __ET_UNUSED const DataLoader::SegmentInfo& segment_info) const override {
+    Result<FreeableBuffer> load(size_t offset, size_t size, ET_UNUSED const DataLoader::SegmentInfo& segment_info) const override {
         NSData *subdata = [data_ subdataWithRange:NSMakeRange(offset, size)];
         return FreeableBuffer(subdata.bytes, size, nullptr);
     }

examples/arm/executor_runner/arm_executor_runner.cpp

@@ -52,7 +52,7 @@ unsigned char __attribute__((
 
 void et_pal_init(void) {}
 
-__ET_NORETURN void et_pal_abort(void) {
+ET_NORETURN void et_pal_abort(void) {
 #ifndef SEMIHOSTING
   __builtin_trap();
 #else
@@ -64,13 +64,13 @@ __ET_NORETURN void et_pal_abort(void) {
  * Emit a log message via platform output (serial port, console, etc).
  */
 void et_pal_emit_log_message(
-    __ET_UNUSED et_timestamp_t timestamp,
+    ET_UNUSED et_timestamp_t timestamp,
     et_pal_log_level_t level,
     const char* filename,
-    __ET_UNUSED const char* function,
+    ET_UNUSED const char* function,
     size_t line,
     const char* message,
-    __ET_UNUSED size_t length) {
+    ET_UNUSED size_t length) {
   fprintf(stderr, "%c executorch:%s:%zu] %s\n", level, filename, line, message);
 }
 

exir/backend/test/demos/rpc/ExecutorBackend.cpp

@@ -47,7 +47,7 @@ class ExecutorBackend final : public PyTorchBackendInterface {
   Result<DelegateHandle*> init(
       BackendInitContext& context,
       FreeableBuffer* processed,
-      __ET_UNUSED ArrayRef<CompileSpec> compile_specs) const override {
+      ET_UNUSED ArrayRef<CompileSpec> compile_specs) const override {
     // `processed` contains an executorch program. Wrap it in a DataLoader that
     // will return the data directly without copying it.
     MemoryAllocator* runtime_allocator = context.get_runtime_allocator();
@@ -129,7 +129,7 @@ class ExecutorBackend final : public PyTorchBackendInterface {
   }
 
   Error execute(
-      __ET_UNUSED BackendExecutionContext& context,
+      ET_UNUSED BackendExecutionContext& context,
       DelegateHandle* handle,
       EValue** args) const override {
     Method* client_method = static_cast<Method*>(handle);

extension/apple/ExecuTorch/Exported/ExecuTorchLog.mm

@@ -116,10 +116,10 @@ - (void)logWithLevel:(ExecuTorchLogLevel)level
 void et_pal_emit_log_message(et_timestamp_t timestamp,
                              et_pal_log_level_t level,
                              const char *__nonnull filename,
-                             __ET_UNUSED const char *function,
+                             ET_UNUSED const char *function,
                              size_t line,
                              const char *__nonnull message,
-                             __ET_UNUSED size_t length) {
+                             ET_UNUSED size_t length) {
 #if ET_LOG_ENABLED
   NSTimeInterval timeInterval = timestamp / 1000000000.0;
   NSUInteger totalSeconds = (NSUInteger)timeInterval;